How We Decreased a CPG Beauty Brand’s ACOS from 72.26% to 46.63% (target 50%), while increasing revenue 40.5%

With G-d’s help:

This case study details the tactics we at 011 Ads for Brands used to decrease a beauty client’s Sponsored Brands ACOS from 72.26% to 46.63% (target was 50%), while increasing revenue 40.5% .

If you like this article, get a live audit of your account to see how to increase your Amazon ad sales / decrease ACOS. You might also want to sign up for the case study that shows how we used Store Spotlights to make $425,000 from 2 vague, previously underperforming keywords.

In this post, we’ll talk about :

  1. Scaling Sponsored Brand campaigns with single keyword campaign,
  2. Bulk Operations spreadsheets – hopefully you’ll find them less scary and intimidating after reading this –
  3. Custom Images when “Embarrassing Conditions” or other Amazon policy limitations affect your choice of images
  4. When Store Spotlights don’t work as well as Sponsored Brand Product Collections

Initial Analysis & Strategy:

When we took over this account, we could see that it had been long neglected:.

  • The most recently launched Sponsored Brand campaigns had been made 4 years ago.
  • Keywords with different meanings were mixed together in the same campaigns. Since your Sponsored Brand headline applies to all the keywords, the relevance and performance varied dramatically from one to the next. The ability to adjust bid by placement – which is managed at the campaign level, not the keyword level.
  • Many bids on unbranded (aka generic) keywords were over $6 and many CPCs around $4-$5… for $20-$25 products! That means that even if you had a remarkable 25% conversion rate, you’d still have 100% ACOS!.
  • Broad match targeting wasn’t modified. Unmodified Broad Match lets Amazon show your ad for whatever their algo deems “related.” It’s roughly the Sponsored Brands equivalent of Sponsored Products’ auto campaigns, with a bit of guidance from your keyword. It’s understandable they were using Unmodified Broad Match since this change to broad match was made about 3 years ago (after the last time they made new campaigns). The outcome was a lot of wasted spend on irrelevant search terms – as well as discovering some decent search terms they hadn’t targeted deliberately.

As a result of these issues, ACOS was 72.26% the month before we took over, and had averaged about 79% in the three months prior, on average sales of $33,406.

The client’s goal was a breakeven ACOS of 50%.

First steps: Mass Scale of Single Keyword Sponsored Brand Campaigns & Store Spotlights

We have a custom tool that allows you to build 100s or 1000s of Sponsored Brand campaigns.

We took the existing campaigns’ highest selling keywords (top 20%) and used our tool to recreate the campaigns with just one keyword per campaign. We didn’t change the headline or merchandise for now.

As a result of doing this, we were able to decrease bids on Other Placements for campaigns where the Other Placement performance was poor. We couldn’t do that before, because multiple keywords in a single campaign means that your placement adjustment will apply to all of them. And they almost certainly won’t have identical performance across placements.

When a new campaign launched, we tried to smooth the transition by letting the two campaigns run in parallel for a few days. Then we paused the keywords from the original multi-keyword campaigns.

In parallel, we made Store Spotlight campaigns to target some of their more niche terms. EG “Main keyword for people with x kind of skin.” We were surprised to find that, although our clickthrough rates went up – as usual with Store Spotlight campaigns – our conversion rates went down.

Note: We didn’t do a clean test. Unlike the previous point, here we did change the merchandise and headline from the Product Collection Sponsored Brand ad that we were trying to improve on. However I think it had more to do with a general trend we see that Store Spotlight campaigns perform better for top of funnel keywords. They don’t perform as well for bottom of funnel keywords.

If you need to decide between the two, think about how you would segment the audience with different Store Spotlight pages. If you find the difference between pages are not significant, that’s a sign you should use a Product Collection instead.

Summary: We reorganized the campaign structure from a few campaigns with a mix of different keywords to many single keyword campaigns. This gave us the ability to manipulate bids on other placements and more control over the campaign.

Second Step: Bulk Operations

Next, we dealt with “bad” keywords.

We downloaded the Bulk Operations file for the last 60 days.

After adding filters, we easily found the worst-performing keywords. Those that were had high spend and low sales. So we changed the status of those to paused.

This links to a common question. “How much data do you need to make a change such as pausing a keyword or altering a bid?”

The answer is that you need enough data to see a pattern.

The Talmud has a discussion of what constitutes a pattern, with a debate as to whether two data points or three is the minimum. Everyone agrees three is a pattern, but what about two data points?

Based on that and leaning towards being conservative, we make changes after seeing “enough clicks for four data points,” i.e. four conversions. If you have a 10% conversion rate, that means 40 clicks. Less than that, and we let it run longer.

Interestingly, we once had a freelance math PhD run the numbers on how many clicks it would take to confirm the null hypothesis with 95% (maybe it was 98%?) certainty, for a new keyword assuming a conversion rate of 10% on the account. In other words, how many clicks does it take before we can say this keyword is almost certainly not going to convert, and we should pause it? His answer: 35 clicks .

Note that this account converted on average a bit higher, around 12-15%. So for them we could make decisions on less than 40 clicks.

In parallel, we took the opportunity to decrease the massively-overbid keywords in bulk. We converted the bid column from text to numbers (Excel offers this as a suggestion when you select a string that consists of numbers). Then we sorted the Max Bid column high-to-low. After adding in a filter to see high ACOS rows only, we adjusted bids down to more reasonable levels, on those terms that were relevant but just overbid.

After uploading this sheet, we saw an immediate improvement. ACOS went from the 72-75% range to the low 60s and continued to improve into the 50s as time passed. Sales were largely unaffected as we mostly eliminated poor performing keywords and made bids more realistic.

Summary: After filtering through Bulk Operations and pausing poor performing keywords, ACOS decreased dramatically and continued to decrease over time.

Third Step: Expanding Match Types & Keywords Plus Adding Custom Images

Next, we doubled-down on keywords that were working well by targeting them in additional match types as well as testing Custom Images.

Some of the original advertising just promoted products using phrase match or broad match. We added exact match and broad match modified where it made sense.

For those who are wondering, broad match modified restricts broad match to include your targeted keywords in the search term. You do this by adding plus signs right before each word you want to appear in the search term. So instead of targeting e.g. skin cream, you would target +skin +cream.

Something else we did (hat tip to IntentWise for the idea) was to look at what search terms were successful in Sponsored Products and check which ones weren’t being targeted in Sponsored Brands. That turned up some gems that contributed a significant amount to increasing sales.

Here’s how to do this on your own.

  1. Download your Sponsored Products and Sponsored Brands search term reports.
  2. Copy paste each of them into a separate tab in a new spreadsheet.
  3. Sort each search term report by impressions high-to-low.
  4. Make a list of search terms by copy-pasting both search term columns into a third tab.
  5. Use Excel’s deduplicate function to eliminate duplicates.
  6. Write a vlookup formula in Excel to find impressions for each search term in one report and write a second vlookup to find impressions for that search term in the other report.
  7. Use conditional formatting to highlight cells where there are 0 or very few impressions.
  8. Divide one impression column by the second to see the difference in percentages.Note that Sponsored Products normally will show lots more impressions due to keyword-targeted SP campaigns also showing up on product pages. Sponsored Brands don’t have this horrible “impression bundling” of search and display so you see fewer impressions there.

Custom Images

Unfortunately, adding custom images isn’t as straight-forward as some experts would have you believe. While many custom images do give you some increase in clickthrough and/or conversion rates, others hurt your clickthrough and conversion rates. So you need to test and track this, not just upload these with a “set and forget” attitude.

In our case, this was even more challenging as the product line we were promoting helps to address a condition. Amazon Advertising’s rules prohibit custom images that show “embarassing” conditions. So Amazon wouldn’t allow us to use before/after shots in the Custom Image, which would be ideal to show their effects.

Instead, we tried some plain jane product photos, which didn’t do much for performance. Eventually we found an “after” photo that was decent and did help boost performance, thanks to persistence in a/b testing.

Summary: Expanding to additional match types, ab testing custom images and comparing Sponsored Brands against previous Sponsored Products campaigns added to the increase in sales. These came at a low ACOS, improving the account average.


The account initially had about 7 Sponsored Brands campaigns, which we built out to over 200. While there were some small wins from Placement bid adjustments, this didn’t move the needle for us initially. Later on it played an important role by making a/b testing easier, by isolating the keyword we were testing.

Working with bulk operations sheets allowed us to quickly pinpoint the worst-performing keywords, as well as those whose bids were too high. Finding them was just a matter of adding filters to our headers, then playing around with different thresholds of ACOS and spend to identify the underperformers. By pausing the poor performers and reducing bids on relevant but overbid keywords, we took major steps to getting the ACOS under control.

Finally, expanding targeting allowed us to generate incremental sales at a low ACOS. We found these additional keywords based on semantic relationships, expanding match types and learning from Sponsored Products. This brought the ACOS from the mid 50s into the mid 40s – slightly profitable and meeting the client’s 50% goal. At the same time, sales increased to a peak of $52,666, increasing revenue by over 40%.

Post script: Unfortunately, due to a huge blunder by Amazon, the client got kicked off Amazon part way through October. Hence the chart cutting out there.

If you liked this article:

You might also want to check out how we used Store Spotlight Sponsored Brands to make $425,000 from 2 vague, previously underperforming keywords.